Automated Disease Detection in Silkworms Using Machine Learning Techniques

Silkworm diseases pose a major threat to the sericulture industry, with early detection remaining a challenge due to limited infrastructure. This study focuses on detecting Grasserie disease, which can rapidly spread in silkworm rearing units, leading to significant economic losses. A novel dataset of 668 healthy and 574 Grasserie-affected silkworm images forms the basis of this research. The study applies machine learning techniques, using the Histogram Oriented Gradient (HOG) feature descriptor combined with Kernel Principal Component Analysis (KPCA) and supervised classifiers. The integration of Support Vector Machines (SVM) with HOG and KPCA achieved high accuracy (93.16%), recall (93.38%), and precision (91.94%), offering a faster, more accurate alternative to manual detection methods. This approach holds great potential for developing real-time, IoT-based diagnostic tools that enable farmers to quickly identify infected silkworms, reducing disease spread and economic losses, and can be extended to other agricultural applications requiring early disease detection.